Skip to main content

The possibility of automatic generation of designer materials

· 2 min read

TLDR

In this article, I will share my thoughts on the automatic generation of designer materials based on my experience working on a Design Platform project. I'll discuss how we implemented a semi-automated system for creating marketing visuals, the principles behind it, and explore the broader implications for design automation. This approach demonstrates how combining designer-provided frameworks with rule-based generation can create scalable design solutions while maintaining quality standards.

In 2020, after graduating with my bachelor's degree, I participated in my first project, which was Design Platform. One of the most beneficial functional modules was the Automatic Generation of Marketing Visuals. At that time, many large companies in China had similar functionalities.

The principle of the solution we adopted was actually not complicated:

  • Designers provided a content framework that met aesthetic standards through Visual Building.
  • Operation staff filled the marketing copy.
  • Illustrations, background-image were generated based on the aesthetic rules given by the designers.
  • Multiple results were displayed on the webpage through permutation and combination of materials, and the operation staff browsed through them one by one to select the best one for publication.

It is evident that this is essentially a production chain, where the product is continuously stacked and completed along the processing chain, finally passing quality inspection and being releasing.

One interesting point is that the number of materials generated by rule-based permutations and combinations far exceeded expectations.

In fact, this model can be applied to any scenario where the products have similarities, such as landing pages, event pages, banners, posters, charts, forms, and other common scenarios.

Only two core elements are needed:

  • 1.Sufficient materials: The increase of materials at each link has an exponential impact on the final result.
  • 2.Reasonable rules: Ensure that the final product is not randomly pieced together.